AMD Looks To 'Crush' Intel's Xeon With New Epyc Server Chips (extremetech.com) 136
AMD has unveiled the first generation of Epyc, its new range of server processors built around its Zen architecture. Processors will range from the Epyc 7251 -- an eight-core, 16-thread chip running at 2.1 to 2.9GHz in a 120W power envelope -- up to the Epyc 7601: a 32-core, 64-thread monster running at 2.2 to 3.2GHz, with a 180W design power. From a report: These chips are built on the same fundamental architecture as the company's Ryzen CPU cores, and they're aimed at the incredibly powerful data center market. AMD's 32-core / 64-thread Epyc CPUs combine four eight-core dies, each connected to the other via the company's Infinity Fabric. According to AMD, this approach is significantly cheaper than trying to pack 32 cores into a single monolithic die -- that approach would leave the company potentially throwing away huge amounts of silicon during its production ramp. The Infinity Fabric is deliberately over-provisioned to minimize any problems with non-NUMA aware software, according to Anandtech. Each 32-core Epyc CPU will support eight memory channels and two DIMMs per channel, for a total maximum memory capacity of 2TB per socket, or 4TB of RAM in a two-socket system. Each CPU will also offer 128 lanes of PCI Express 3.0 support -- enough to connect up to six GPUs at x16 each with room left over for I/O support. That's in a one-socket system, mind you. In a two-socket system, the total number of available PCI Express 3.0 lanes is unchanged, at 128 (64 PCIe 3.0 lanes are used to handle CPU -- CPU communication). Anandtech has a longer writeup with more details on the CPUs power efficiency and TDP scaling. Further reading: ZDNet, press release.
Hopefully not too late (Score:4, Interesting)
Re:Hopefully not too late (Score:5, Insightful)
Who knows what Intel might have in their back pocket after chilling around for half a decade?
Haven't we already seen some indication that Intel's been caught flat footed? Their response to Ryzen has been strange and erratic with confusing product launches (or at least announcements...) the tech media is still trying to untangle, $1500+ 175W i9s that hardly anyone will buy and other bad ideas.
I expect their marketing people will sperg out and invent some bizarre new name for embiggened chips based on existing process tech and blow a few million on the "launch event."
The fact is AMD has caught up. So now there is competition and Intel has to compete. If Intel does have anything "in their back pocket" they'll have to use it to hold the market share they've earned. If not then they'll have to compete on price etc. It's all good. We win either way.
Re: (Score:2)
AMD may be competing on cost but it hasn't caught up on performance, power consumption nor compatibility. Intel is still king when it comes to servers, it might be good for gamers where interconnects between CPU's, memory and massively parallel IO is not as important but I guess I'll have to see what the current crop of AMD processors brings.
Re: (Score:2)
Cost generally is for a specific level of performance and heat. It is like Ohm's Law, you can't separate the parts. Different sides of a die.
As far as compatibility, you have to buy a motherboard for one or the other, and everything else works the same. That's a complete wash, and has been since the EISA bus went out of fashion.
Re: (Score:2)
When you fill a datacenter with machines, the price and performance per unit does matter. The internals of the CPU and how it connects to the systems don't quite work the same (last few iterations Intel has had more bandwidth between CPUs as well as between CPU and high-speed peripherals). AES for example is about twice as fast on Intel as AMD and the virtualization set isn't exactly the same leading to Intel support being earlier and better supported.
Re: (Score:2)
Reading comprehension matters, too.
Also, not very much in a datacenter is encrypted. Intel has enhanced AES because they have enterprise remote management that has be encrypted and runs outside the main CPU cores. So it is more built-out.
Re: (Score:1)
Re: (Score:2)
Judging by their knee-jerk response, no they don't.
Re: (Score:1)
Re:Hopefully not too late (Score:4, Insightful)
Yes, but only to maintain, but not eliminate AMD.
You see, Intel's goal is to keep AMD where they are. They WANT AMD to have a foothold in the server market where the money is. But not too big a foothold.
For Intel, AMD"s position is perfect - they are the distant second and will nip at the heels like an eager puppy. Which is where Intel wants them. Too powerful and they have competition, too weak and AMD could die off and unleash all sorts of government investigations, regulations, anti-monopoly rulings and other things. Worse yet, AMD's patents that are cross-licensed with Intel might go to ARM or others forcing Intel to pay or cross-license.
So Intel needs AMD to be around, but not too powerful nor too weak. They could crush AMD with what they have, but they won't, so they just need to deploy "just enough" to counter the threat without actually harming AMD.
Re: (Score:2)
Don't know where you got that impression.
Intel already showed its hand with the i9 series they shoved out the door as a knee-jerk reaction to AMD. It's a mess.
They've basically taken existing Xeon cores and prepped them for consumer use -- but with crazy options for the sockets for the new pin counts needed -- making motherboard manufacturers have to deal with different CPUs all for the same slot type, but some have different feature sets meaning it's hard to make a mobo that works with all of the chips t
Re: (Score:2)
Don't know where you got that impression.
Intel already showed its hand with the i9 series they shoved out the door as a knee-jerk reaction to AMD. It's a mess.
They've basically taken existing Xeon cores and prepped them for consumer use -- but with crazy options for the sockets for the new pin counts needed -- making motherboard manufacturers have to deal with different CPUs all for the same slot type, but some have different feature sets meaning it's hard to make a mobo that works with all of the chips that fit that slot type.
Most of the i9 series is vaporware -- sure, they have the specs laid out, but they only have a butchered Xeon to fit the new slot type to test with so far.
AMD is ready to roll out and has chips available today for Ryzen and soon Epyc. Meanwhile, Intel likely will take a year or more to offer i9s. Sure, Intel will eventually switch production lines and tailor the chips better and come out on top eventually... but, they were caught with their pants down on this.
Intel's solution is basically to offer the Xeons as consumer-level products with more cores and a different pin count (but without ECC ram compatibility) to match the higher level Ryzen offerings... and now they'll have to do even better to match the Epyc offerings at the server-level. Most likely, it'll take them 2 years to repackage the Xeons to take advantage of as many PCIe lanes as they'll likely have to change the slot type again and work with mobo makers to get the right thermal envelope down for the new slot type on their boards. The PCIe lanes are the key difference as AMD is betting a lot of AI and graphics rendering is done on GPUs... and their Epyc chips will let companies connect a lot more GPUs and other peripherals at high speed by PCIe to the multi-core CPU. Intel..... doesn't currently have anything like that -- and it will take them time just to re-purpose a current chip to support that, much less roll it out in large production.
AMD runs at 70% of the power requirements of Intel's compatible device and at 30 percent price reduction. Benchmarks show AMD's offering is within 5% of Intel's non-pushed (standard chip) performance.
For 30% reduction in electricity costs, coupled with less than 5% difference in performance, the decision process is easy to make.
Re: (Score:3)
Intel very likely has nothing. They made huge profits being fat and lazy and they expected that to continue. Remember that comparable small AMD has declassed Intel before. Intel may well need several years to catch up.
intel has cut down pci-e and held back boards that (Score:2)
intel has cut down pci-e and held back boards that have to work with cpu's that are just desktop cpus stapled on an workstation / 1 cpu server layout / chip set.
So an board with up to 4 ram channels and up to 44 pci-e lanes from cpu have to work with cpu's that only have 2 ram channels and 16 pci-e. Maybe some will just feed slots from the X4 DMI feed chipset to give them more io over lot's of different cpus.
So it's not a 32 core chip (Score:5, Informative)
It's a cluster of 4 8 core CPUs connected via a high speed interconnect. I'm not saying that is bad; I just wish a tech site would have more accurate reporting.
Re:So it's not a 32 core chip (Score:4, Interesting)
Re: (Score:3)
Too bad you couldn't afford to patent it....IF you were going to do something practical with it. That I have no problem with.
If not, I'm glad you didn't. That's the problem with patenting "ideas"....a lot fail to implement them, then stifle the progress of the industry because it's easier to litigate than innovate. There's a whole thriving industry around that, which contributes nothing to society.
I swear....if there was a short time limit on how long a patented "idea" had until implementation before bei
Re:So it's not a 32 core chip (Score:5, Informative)
Re: So it's not a 32 core chip (Score:1)
It's always Intel versus AMD. It's never about the good that having choices offers to us.
At least, I guess that's what is important in 'nerd culture' (oxymoron?)
Re: (Score:2)
This whole conversation has been going back to the mid 2000s when Intel released their quad core CPU that was really two different dies glued together.
The 1995 Pentium Pro was two chips (core and cache) in one package. They called it a Multi-Chip Module.
Re: (Score:3)
You need to take into account that AMD CPU interconnect and thread migration always was far, far superior to Intel. Usually you lose low single digits in performance on AMD, while on Intel people hat to use an additional system to stream gaming, for example, because Intel inter-CPU communication is so bad.
Re: (Score:2)
It's a cluster of 4 8 core CPUs connected via a high speed interconnect. I'm not saying that is bad; I just wish a tech site would have more accurate reporting.
As if "core" has any well-defined meaning besides marketing slogans. IMHO, it has 32 CPUs within a single package, in groups of 8 per die. Operating systems and applications will see 64 CPUs due to SMT.
I remember when the "multicore" chips came out around 2005, and it was portrayed as something revolutionary and hard to program efficiently. I had already been coding on multi-CPU systems years before, and putting the different CPUs on the same die didn't seem like a big deal. In fact, the shared caches an
Epyc? (Score:3, Interesting)
How do you even pronounce Epyc? Like "epic"?
Re: (Score:2)
Re: (Score:3, Funny)
Or I just say that for a laugh because yeah they're going for "epic".
Re: (Score:2)
I like your first idea better and I'm adopting it for myself . Thank you!
Re: (Score:2)
Re: (Score:2)
Yes! Its fucking awesome, Hopefully they keep rolling out these power packed processors. Hope Vega is what theyre claiming.. nvidia needs a kick in the dick too.. even though i have a gtx1070 in this pc.
Re: (Score:2)
My first guess was 'E-Puke'
Re: (Score:1)
he 'pike' bit is about the pike or pickerel, a largish long freshwater fish,
Actually I like the idea of it being named after the fish. I used to fish for pike. I'd have to use steel leaders on my line as pike have some vicious teeth. I've seen them eat ducks and a small dog once. Most of the time they would bite a duck in half with the first bite and then come back for the other half on the second pass. They're also strong swimmers and fight like hell when you hook them.
Re: (Score:2)
The "pyc" is silent.
Re: (Score:3)
The "pyc" is silent.
So these new AMD processors run Python compiled bytecode? Sweet!
https://stackoverflow.com/questions/8822335/what-do-the-python-file-extensions-pyc-pyd-pyo-stand-for [stackoverflow.com]
Re: (Score:2)
As opposed to literally every other x86 chip on the market (and many others, besides), which also run python compiled bytecode?
Wooosh!
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
Goofy unpronounceable names dissuade me from buying a product.
Re: (Score:2)
So does this mean AMD are good again?
No.. It means they still ARE good (tongue firmly in cheek) ... Only now they are getting competitive on their power consumption vrs processing power numbers...
My problem with AMD (Score:1)
Re: (Score:1)
About ten seconds of searching on Newegg found dozens of AMD systems. Same on Amazon. It's not AMD's fault that you're bad at simple searches.
Re:My problem with AMD (Score:5, Informative)
https://www.extremetech.com/computing/184323-intel-stuck-with-1-45-billion-fine-in-europe-for-unfair-and-damaging-practices-against-amd
AMD wasn't exactly partying like they ruled the world, the deck was stacked against them by Intel when AMD had a superior product.
Re: (Score:1)
Among them having CEO who printed posters with himself styled as Indiana Jones protecting his trophy wife.
Yes, these were posted around the Austin campus.
Re:My problem with AMD (Score:5, Informative)
The reason AMD couldn't sustain is that Intel bribed their customers to stay with them as shown by the lawsuit Intel lost and the 1.2 billion dollars that Intel owes AMD. Intel has made many times that by essentially marginalizing AMD so the fine is most likely worth it to them. It is worthy to note that AMD still hasn't received a penny of that fine from Intel since they agreed to pay in 2009.
There was a separate, similar case in the UK with a similar outcome. Intel owes another 1 billion euros to AMD from the second case which it has been appealing unsuccessfully for several years.
Overall Intel owes AMD close to 3 billion dollars in fines but has successfully delayed paying it for the better part of a decade.
Re:My problem with AMD (Score:5, Informative)
Can you buy it? AMD arguably made "ok" CPUs (talking within past 2-3 years), and even some mildly interesting ones. But could you buy them anywhere?
Yes [newegg.com] you [amazon.com] can. [google.com]
Re: (Score:2)
https://www.supermicro.com/pro... [supermicro.com]
https://blog.dellemc.com/en-us... [dellemc.com]
https://www.forbes.com/sites/p... [forbes.com]
They sure have AMD now (or coming shortly).
Re:My problem with AMD (Score:5, Insightful)
Also, the last time AMD put a dent into Intel, Intel started fighting back in a large number of ways that were later found to be illegal [theverge.com]. Celebrate and relax, indeed.
Re: (Score:2)
I can understand why board makers weren't very interested in supporting AMD's second-rate CPUs the past few years. It's nice to see them offering something competitive again.
Re: (Score:2)
Can you buy it? AMD arguably made "ok" CPUs (talking within past 2-3 years), and even some mildly interesting ones. But could you buy them anywhere?
You are kidding right?
I have 4 retail establishments within 30 min of my home which have AMD CPU's available for immediate purchase, usually with motherboards and all the rest of the stuff you need to build a system. Then there is Google (assuming it's not broken where you are) where scads of mail order suppliers are easily found.
Now if you mean "available in pre-build solutions", then I suggest you take a closer look at what that means. Intel has huge power to strong arm system integrators into using i
Re: (Score:2)
I'm pretty sure he was talking of ready to use AMD servers, which have been nowhere to be found for the last 5 years or so.
Re: My problem with AMD (Score:1)
Can't the server companies buy AMD motherboards and lots of Phillips screwdriver and... and... and...
Re: (Score:2)
Plenty of servers on Pricewatch.com
I'm surprised someone of your UID doesn't know of the place. Newegg advertises there and gets beaten fairly regularly.
Re: (Score:1)
Oh yes, you can find old servers / much less powerful ones on the web, but that`s not the kind of server that gets purchased repeatedly. We could always have replacement on hand in our small market (Quebec) for intel servers, or upgrades to the newer ones, even rare raid unlock chips, but there was nothing on hand for AMDs. With a 99-1 split in the server market, that is expected I guess..
For some nice read, the following & further;
http://marketrealist.com/2017/... [marketrealist.com]
Re: (Score:2)
Uh, no. Pricewatch already has Ryzen out and as soon as EPYC hits the market they will have it as well.
Not my fault you're in Quebec where if the product doesn't have fucking French on it then it can't legally be sold there. Yes, I've fucking been there. That was 17 years ago. Haven't gone back since because of that intolerant bullshit.
Re: (Score:1)
Ryzen is not a server cpu, and it's already available here too.
Re: (Score:2)
Ryzen is just a cut-down EPYC chip, so yes, it is a server CPU. Even supports ECC.
Re: (Score:1)
Supporting ECC doesn't make it a server cpu. Similarly, using ECC memory doens't make a computer a server -- see the late mac pro -- so much trouble running a farm of those! I would not be running ryzen chips in my 2U enclosures -- performance / volume would not be good at all. Anyway I am no longer working with local servers.. While I am eagerly awaiting the 128 cores epycs to see how they fares, I will simply start using them should the price / performance ratio becomes better than corresponding intel ser
Re: (Score:2)
"Supporting ECC doesn't make it a server cpu."
In reality, that's about the only feature separating consumer and server-grade chips now days. Everything else is either feature-compatible, or cut from the same die and just laser-locked.
Re: (Score:2)
Re: My problem with AMD (Score:1)
Re: (Score:2)
It isn't a question of AMD slacking off. Intel has always had a process advantage, but as physical limits are approached that advantage is diminishing. The only time AMD outperformed Intel was during Intel's netburst-RAMBUS blunder.
Intel has a financial advantage that allows it to outspend AMD 10:1 on research. As the process advantage slowly becomes history, the research advantage that gives Intel more cycles-per-clock is all that's going to be left. I think that also will diminish over time as X86 archite
Re: (Score:2)
Re: (Score:2)
At the current high density of about 400 Xeon per rack (could be higher this is the most I've seen) your still looking at 250 racks filled with blade servers alone, Somewhere your going to need network and storage nodes as well. Those cool with air just fine.
Re: (Score:2)
Re: (Score:2)
"Can simple air flow cooling keep the data center from burning up?"
I can cool 300w of LED on a LGA775 heatsink and keep it at 70C.
So maybe you might want to think about the cooling solution instead of the processor itself, n00b.
180 Watts (Score:1)
Re: (Score:2)
Yeah this! I thought we were past the days of these super hot TDPs. The hottest Intel Xeon is 155W I believe. Is this going to be so desirable in a datacentre of all places where power consumption and cooling are already a significant portion of a system's ownership costs?
Re: (Score:2)
The answer is yes.
Because the critical considerations are compute efficiency (wallop per watt) and compute density.
Escalating wallop density is ultimately going to cost you on the TDP front, except as compared to free lunch.
A = 2 x (1 platform + 1 CPU)
B = 1 x (1 platform + 1 turbo CPU)
Suppose turbo CPU is 50% hotter, 100% more powerful.
Or perhaps
Re: (Score:2)
Not at all. But I also haven't swallowed marketing hype yet. Let's see just how much faster this is in real terms before we go accusing each other of popping mythical pills.
whay that is (Score:2)
There are at least two forces driving this, one physics, the other economic:
The big picture is that, as feature size approaches physical limits of silicon, progress toward ever-higher gate density has slowed. In fact, Intel has had to revise their Tic Tock design cadence to pad in more CPU architecture re-designs as they await finer processes. So though Intel was way out ahead, their rate of progress slowed, allowing AMD to catch up.
The other point is, Intel's positive re-enforcement loop of having the b
Re:whay that is (Score:4, Interesting)
The other point is, Intel's positive re-enforcement loop of having the best processors because they had the most revenues from selling the best processors to develop new processor fabs has been broken by ARM. Competition from ARM architecture drives fab R&D outside of Intel, so much so that Samsung foundries were shipping Qualcomm ARM processors at 10nm while Intel was still stuck at 14nm.
No, Intel just dropped the ball on that (or decided to let go of it). They had an 18-24 month lead and they let TSMC and Apple catch up.
Maybe they decided that dragging out (prolonging the inevitable end of) scaling by 4 years or so would be more profitable between now and 2025 than galloping to the finish line.
Re: (Score:2)
Part of the consent decree is that AMD is intitled to use any/all related Intel patents. AMD is allowed to put AES into microcode. I would have preferred that they put twofish or threefish instructions.
The latter two are stonger than AES for hacking and are fast in execution. But being much harder to hack than AES, you can guess why blowfish/twofish was not chosen.
Re: (Score:2)
The problem is that moving to a new CPU architecture is something that not many companies want to do. POWER and SPARC are all but dead, which pretty much leaves us with x86 and ARM as what runs everything out there. I will say the Itanium was a definite step up, with hundreds of registers, but getting Linux ported, and a top tier distro maker to actually port packages to that architecture, and support it, will be an uphill battle.
The exception would be an architecture that can virtualize x86, allowing sof
Re: (Score:1)
10nm and 14nm are marketing terms. The actual silicon measurements in each is very much reversed in the specific case you note. That this actually matters or not, I don't know. Numbers must go higher, er lower, whether they mean anything or not.
EPYC vs Broadwell (2.5 yr old CPU architecture) (Score:1)
Data centers and GPUs (Score:2)
Seriously, I just don't know: why would a product targeting data centers, make a big deal about connecting GPUs? Are a lot of them really doing that, or are they using the great I/O bandwidth for something else (what?) or does this usually go greatly underused?
(Yes, I realize there are applications f
Re: (Score:1)
There are also applications for PCI-e that are not GPUs.
Re: (Score:2)
Tell me more. I can think of some things, and yet when I imagine these data center computers (anyone else get it yet? I don't work in a data center!) I mentally picture most of them as having nothing plugged into their PCI-e slots.
Ok, I picture one that has a bunch of SAS adapters plugged into there, and runs a fuckton of disks or SSDs. But then the others just use those disks over the network or storage network.
I maybe picture one that has some amazingly fancy network adapter in it .. and a bunch of othe
Re: (Score:2)
Also quite a few lanes need to be dedicated to high speed network adapters. Modern servers often have at least 2 NICs that run anywhere from 10 GBps up to 100 GBps each.
The other thing high numbers of lanes is used for is compute servers with large numbe
Re: (Score:1)
I said there are applications that were not GPUs.
You were able to come up with some all by yourself.
Why do I need to tell you more? You answered your own question.
Re: (Score:2)
You are wrong. GPUs are being used more and more for general compute. From AI to bitcoin mining they're the way to go. NVidia and AMD both make GPU cards that have no video outputs specifically for data center use.
Re: (Score:2)
Just to name a few server components off the top of my head (and yes, I do this for a living);
* GPU's used for large scale data analysis and AI. Big usage in large datacenters already today, particularly companies where large amounts of data require quick processing (think GIS for example)
* Network interfaces. The next generation of networking will bring 25Gbps Ethernet as well as bring greater support for 100Gbps Ethernet. Throw 32Gbps Fiber-Channel in there, too.
* FPGA's. Possibly to me more exciting than
Question on package (Score:1)
Can someone with knowledge on the topic post what the packaging and cooling strategy for the 175W chip looks like? Is air cooling still going to be enough? Will a typical system still be able to run at 50C ambient?
Re: (Score:1)
That power isn't that far off from high end Pentium4s. Yeah, huge heat-sink and a big fan would do it.
More cores isn't better (Score:2)
Given the same aggregate performance, the fewer cores a CPU has, the better because the less you have to worry about concurrency and cache issues.
More cores _can_ be better (Score:1)
Given the same aggregate performance, the fewer cores a CPU has, the better because the less you have to worry about concurrency and cache issues.
You are right, "given the same aggregate performance" that's true. I understand that the value proposition of AMD is _higher_ aggregate performance thru higher core numbers, and probably lower overall price.
Re: More cores _can_ be better (Score:2)
My point is they should talk about performance, not number of cores.
Support? (Score:3)
Is AMD assigning engineers to help port stuff that has gone Intel-only in the years that AMD failed to compete in the server space?
I'm thinking of some of the newer Xen features, codec assembly, etc. If we have to wait a year to try Ryzen they may have missed their opportunity.
Can't wait for Ryzen laptops. (Score:1)
AMD has seemed willing to challenge Intel in their traditional strongholds. Their laptop APUs will be becoming out later in the year and will no doubt be priced aggressively to challenge Intel. They'll probably start with a basic quad-core at the bottom, six-core (with HT) in the middle, and eight-core with hyperthreading at the top.
The Zen-based desktop APUs are also coming out. They're probably geared for corporate buyers/manufacturers since the integrated graphics mean no discrete GPU is needed. (I g
I know they're server-side, but what a laptop! :) (Score:2)
Personally I'd love to have an 8 core chip running in a laptop with something like 128 or 256 GB of RAM. I already do most of my work from an MBP running VMWare Fusion and having two or three desktops running simultaneously. Something like that would put me in heaven.
Hmm... I wonder how far away we are from running a laptop with at least 16 cores, at least 1 TB of RAM, multi-terabyte SSD all with a battery that could last at least 8 hours? Running multiple instances of machines (Linux, Windows, etc.,) for d
Re: (Score:2)
You're a long ways out from having 1TB of RAM in a laptop, only dog knows why you would want that much.
But the 16 core CPU is possible now and multi-terabyte drives are already here.
huge mistake (Score:2)
24 hot-swap U.2 NVMe drive bays supermicro (Score:3)
24 hot-swap U.2 NVMe drive bays supermicro
now just think if each one had its' own X4 link and you still have PCI-E leftover for say 4 10G links.
Re: (Score:1)
You had promised not to upgrade your computer until AMD was the king of performances again ? I hope that those processors pan out, and I'd say you should get something like the i7-7700k, perfect performances in D2016!
Re: (Score:2)
APK doesn't even know that AMD has always had better 'process scheduler thrashing handling' so you already know his "AV Solution" is utter shit in the face of today's OSes and threats.